Hisnlabs Unveils tajweed-v2: A Breakthrough in Quranic Recitation Technology

Hisnlabs, an innovative AI research organization, has made a significant leap in the field of Quranic recitation technology with the introduction of tajweed-v2. Published on November 6, 2026, this model is designed to enhance the learning and practice of Tajweed, the art of Quranic recitation, by providing detailed phonetic feedback and transcription. The model's architecture and performance metrics have been made available on the global model registry, marking a pivotal moment in the intersection of technology and religious education.

The tajweed-v2 model is built with a focus on efficiency and accessibility. It features a compact architecture with 20.6 million parameters, occupying only 23 MB in ONNX int8 format and 83 MB in fp32/safetensors format. This lightweight design allows it to operate effectively on standard CPUs without the need for GPU support or cloud services. The model processes audio input at an impressive speed of 28 milliseconds for a 4-second audio clip and 74 milliseconds for a 10-second clip, making it suitable for real-time applications.

The model employs a rule engine that provides per-word feedback on pronunciation, aligning with the two primary recitation styles, Hafs and Warsh. This feature is particularly beneficial for learners, as it offers immediate corrective feedback, enhancing their understanding of Tajweed rules. The model's performance is underscored by a low phoneme error rate of just 7.0% on the QDAT dataset, which included 159 learners with human labels. Furthermore, it maintains a madd length accuracy of ±1 count 95.9% of the time, ensuring precision in recitation.

The first product utilizing tajweed-v2 is Daghia, an independent search engine developed in the UAE. Daghia is designed as an alternative to major global search engines, focusing specifically on Quranic content. Users can engage with the model by selecting a surah and utilizing the AI Recite feature, which allows them to recite verses while receiving real-time feedback on their pronunciation.

For instance, when reciting Surah Al-Kahf, users can choose to hide the text and recite from memory. The model then analyzes their recitation, marking each word as correct, to review, mistaken, forgotten, or self-corrected. This interactive approach not only aids in memorization but also reinforces the correct pronunciation of Quranic verses, making it an invaluable tool for learners.

The tajweed-v2 model's technical performance is complemented by its user-friendly interface. The AI Recite feature is designed to be intuitive, allowing users to focus on their recitation without the distraction of on-screen text. The feedback provided by the model is available in Arabic, English, and French, catering to a diverse user base and enhancing accessibility.

The model's architecture includes a sophisticated encoder that processes audio at 16 kHz mono, converting recitations into phonetic scripts. This process involves a connectionist temporal classification (CTC) head that generates phonemes and timing information, allowing for precise alignment with the reciter's performance. The feedback mechanism flags various aspects of the recitation, including madd length, ghunnah pronunciation, and other Tajweed rules, ensuring comprehensive support for learners.

The introduction of tajweed-v2 and its application in Daghia represents a significant advancement in Quranic education technology. By providing learners with immediate, actionable feedback, the model fosters a more effective learning environment. It addresses common challenges faced by students of the Quran, such as mispronunciation and lack of personalized guidance.

Moreover, the model's ability to operate independently of high-end hardware or internet connectivity makes it accessible to a broader audience, including those in regions with limited technological infrastructure. This democratization of Quranic education technology could lead to a resurgence in interest and engagement with the Quran among younger generations.

Hisnlabs' tajweed-v2 model is poised to revolutionize the way Quranic recitation is taught and practiced. By leveraging advanced AI technology, it not only enhances the learning experience but also preserves the integrity of Tajweed rules. As more users engage with Daghia and similar applications, the potential for widespread improvement in Quranic recitation skills becomes increasingly tangible. Hisnlabs has set a new standard in the intersection of technology and religious education, paving the way for future innovations in this vital area.